07 August 2006

Media Review Panel For Fairness In Reporting

The takedown of Reuters has been in the news recently via LGF and my previous post on this matter gives a quick summary of what the factual reporting media, both MSM and online, needs to do to ensure integrity of captured images, be they still images or motion images. The era of easily made fraudulent photorealistic models and animation is within the scope of Hollywood as of 5 years ago and soon coming to a desktop near you.

Today the problem is ensuring integrity of captured images by verifying source material and ensuring its integrity so that it does not misrepresent real world events by the necessary editing for publication. Doing things like cropping, rotating and adjusting color levels to give a good depiction across a wide range of non color calibrated displays or for print publication is an art as much as it is a science. We have seen in the news media, that *crowds* of people, up close to a camera, is literally one or two individuals deep, and maybe a total of 10-15 people for a 'mass demonstration'. Thus crowding the camera gives the 'appearance' of a large crowd and news coverage may not show the larger context of that crowd so as to properly let individuals know what it represents. Recently, much has been made of the purported 100,000 individuals that Moqtada al-Sadr has brought out in Baghdad to show support for Hezbollah, but few people report that the *previous* crowd brought out by him for his group a couple of years ago was TWICE that size. Without context one would think that his support was that of a large base, while, in point of fact, it is one that is decreasing over time. So, without context even properly taken images can lie by the absences of surrounding facts.

The problem demonstrated by Reuters in particular but the entirety of the Main Stream Media, is to take partisan reports from warzones and publish them as 'facts' and not 'assertions'. In point of fact, when dealing with a terrorist operation that looks like civilians, telling who is and is *not* a terrorist is a difficult thing. Further, when looking at damaged buildings in a war zone, it is difficult to tease out exactly which damage took place when and then give a good timeline of events. So the problem of gross digital fraud is compounded by selective access and reporting to make things 'appear' to be one way when, in point of fact, they are not. Add in unscrupulous movement of corpses from place to place to 'show' how bad the damage is, add in 'instant rescue response' individuals who happen to be at EVERY site of damage across a large geographical area, and one begins to understand that fraud is not a single event, single occurance act, but can be a systemic act to woo public opinion via falsifying actual events via selective and staged reporting.

This is difficult stuff to tease apart as it covers a broad area of expertise. I will touch on the high points necessary to disentangle such things and put scope and limitation on information and events.

The first thing to understand is that ALL information has a degree of reliability and accuracy to it. Even first hand evidential reports from eye-witnesses are unreliable, as evidence in any criminal court proceedings with multiple eye-witness events. These accounts may be factual and truthful, but have questionable reliability due to emotions and restriction of viewpoint or mental ability to take in the event. In the Intelligence Analysis world, reliability is always couched in the terminology of a report, and very few reports are ever those of being 'absolutely positive' about anything. When taking a group of information of differing reliability and casting them into context, some things just will not fit well and either point out a problem with context, a problem with the information or some missing piece or factor that cannot be determined. All work done needs a reliability and verification of information as a very first step so that all follow-on work is done with that understanding. In this case all digital documents of an image sort need to be taken in along with all metadata as to place, time, camera and individual taking them, and verified that those things are true and that the image can, indeed, be from that camera, at that place, at that time, taken by that person. This requires a broad overview of this evidence by an expert group in digital photography, imagery capture, motion imagery capture and able to cross-verify the types of color space of individual cameras with resultant images. As I had previously pointed out, the Munsell Color Lab at the Rochester Institute of Technology is a prime place were such work for industry, science and academia takes place. Further, they have background in such typific analysis and forensic analysis of documents. In point of fact RIT covers the entire gamut of digital media of all types from satellite imagery to print, and is a prime focal point for the media industry and academia and has a very high level of integrity and respect in that field.

Second, evidence gathering. Events take place at some place at some time, even if they are distributed digital processing networks, the individual processing nodes have space and time components for their processing time. Similarly, real world events have such components and to properly place events in space and time requires the larger overall geographic context of them. A bombed out neighborhood may not have been bombed all at the same time, for example, and so buildings still standing that are later brought down serve as time guideposts for limits of when events could have taken place. To do this requires a specialty known as BDA: Bomb/Blast/Battle Damage Assessment. This is a busy field and covers areas of military work and law enforcement with each having particular specialties and weaknesses. An example for blast assessment is that if it is of unknown origin a military remote sensing analyst may interpret it one way while an FBI investigator with ground level imagery may see it another. BOTH are needed to tease apart such things as: direct battle damage, damage due to indirect sources during battle, post-damage structural failure, IED or VIED use to create damage, and structural demolition via placed charges. That is a broad field and expertise varies across these areas. Retired BDA Analysts from the military of the US Armed Forces may be hard to come by, but not impossible and a look to civilian workers that had worked with such places as NGIC would be prime personnel to do the initial teasing apart of what is and is not actual bomb/missile/battle damage. The FBI Academy Forensics Unit in Quantico, VA offers expertise in skills that cover the rest of this range from demolitions, arson, IED and VIED, and post-explosion damage assessment on building structures. If these sources are lacking then alternate sources from other highly trained military organizations that are up to speed on modern battlefield BDA and National police/security services that do this work would also be prime areas. Again, retired individuals who are no longer primarily associated with this work would have the requisites of background, experience and independence necessary to do this sort of analysis.

Third, while only in-line, but FIRST in conception, is creating the geospace for these events and being able to march through them time-wise in that geospace. This requires actually having a digital overview of the broader region and then putting in analysis points that are time-delimited. As an example using Qana: one would have the general geospace of that town, buildings and structures from either city maps/graphics or from georectified satellite imagery, and then each individual photo given a time stamp and placement along with attached metadata of who took it and who or what is within it. This is known as geospatial intelligence analysis, and is done by the National Geospatial-Intelligence Agency within the US and it has purview for all things outside the US for this work. Other Agencies do more narrow work, but broad analysis of an area and what is happening in it is done for the military by NGIC and NGA. They are, I expect, busy people, but providing wholly unclassified documents for this region in digital format is something they do for the maritime industry based in the US and they have similar capability for the State Department and Justice Department. Getting analyst time from them is impossible, but a retired analyst from that Agency or one of its contractors may be within the realm of possibility. Finding users that can compile a good GIS based timeline and place images and such in context will be relatively easy as nearly every municipality on the planet has such.

Fourth, analysis of individuals, bomb damage, and other things in context over time. Here the FBI and standard police forensics capability is a prime target for putting all of this information with delimited and imperfect accuracy to use in building the entire structure of events. These individuals will be the ones from the other areas helping to place individual photos and motion images from everything from Israeli guncams to tourist photos into the broader picture so that individual damage assessments and tracking of players can be performed. This analysis may require particular pieces of expertise from a direct color space analysis to software manipulation analysis from graphics arts experts to actual modeling and building of 3D models to look at how original buildings and structures appeared before and after events and to see if anything that should be there is missing or if something has been added. This could even require a virtual mock-up of an entire building or street or neighborhood, at least for exteriors of buildings, which is not too difficult. Many 3D game engines offer this like Doom or Quake, which has a vibrant modelers community to draw upon with many pre-made structures available, as well as the fully interactive Virtual Worlds for user made constructs, like Second Life, which allow a direct 1:1 modeling capability for *everything* by skilled designers and allows real-time virtual fly-throughs to examine various objects and their relationships to each other.

Fifth, is the final panel of noted experts in the various digital media, graphic arts and associated military and law enforcement areas. This problem of using fraudulent or tampered documents to sway public opinion is one that will not go away and will get worse over time. Unfortunately this year's SIGGRAPH is over, and would have served as a *perfect* focal point for all of these communities as they are ALL involved in it. That said, going through that organization to reach the acknowledged international and unbiased experts in these realms so as to convene a panel on this may be the key on this. No partisan panel of scattered individuals with expertise will do the job: only a cross skilled panel used to working within these realms and able to work *together* can put forward the necessary changes to digital based reporting necessary to create a new framework for that reporting. Media organizations and individuals will have an onus placed upon them by the doubts engendered in the public that reporting a *story* is no longer reporting *news*. Things that are *news* may not have a story line, while *stories* may not be things that are immediate news and are only built up over time.

In the coming era of digital media, when the cost of recreating and re-casting movies, stories and other things is no longer the realm of Hollywood, but is a desktop phenomena, the ability to create *realistic* looking events will blur the line of what is and is not *real*. The line between craftsmanship in presenting a good representation of an image and artistry in falsifying an image in whole or in part is the tip of a much, much, much larger iceberg that is surfacing due to modern technology. The framework for detecting fraud and insuring the veracity of such images and video streams is paramount as both of those are already falling into the desktop creation realm. The main answer is for all news and fact based reporting individuals and organizations is to *show their work* from raw images and video streams to final product. That may not be possible in all instances, but for most everything done today as reporting, the demonstration of openness and transparency by those media reporters in making ALL of their video and still images available for review is paramount. Cutting and pasting and taking images out of context to support a *story* is not reporting *news* but creating something wholly other in its doing.

No comments: